Statistics Learning and Universal Grammar: Modeling Word Segmentation

Total Page:16

File Type:pdf, Size:1020Kb

Statistics Learning and Universal Grammar: Modeling Word Segmentation Statistics Learning and Universal Grammar: Modeling Word Segmentation Timothy Gambell Charles Yang 59 Bishop Street Department of Linguistics, Yale University New Haven, CT 06511 New Haven, CT 06511 USA USA [email protected] [email protected] Abstract vision of labor between endowment and learning± plainly, things that are built in needn't be learned, This paper describes a computational model of word and things that can be garnered from experience segmentation and presents simulation results on re- needn't be built in. alistic acquisition In particular, we explore the ca- The important paper of Saffran, Aslin, & New- pacity and limitations of statistical learning mecha- port [8] on statistical learning (SL), suggests that nisms that have recently gained prominence in cog- children may be powerful learners after all. Very nitive psychology and linguistics. young infants can exploit transitional probabilities between syllables for the task of word segmenta- 1 Introduction tion, with only minimum exposure to an arti®cial Two facts about language learning are indisputable. language. Subsequent work has demonstrated SL First, only a human baby, but not her pet kitten, can in other domains including arti®cial grammar learn- learn a language. It is clear, then, that there must ing [9], music [10], vision [11], as well as in other be some element in our biology that accounts for species [12]. This then raises the possibility of this unique ability. Chomsky's Universal Grammar learning as an alternative to the innate endowment (UG), an innate form of knowledge speci®c to lan- of linguistic knowledge [13]. guage, is an account of what this ability is. This po- We believe that the computational modeling of sition gains support from formal learning theory [1- psychological processes, with special attention to 3], which sharpens the logical conclusion [4,5] that concrete mechanisms and quantitative evaluations, no (realistically ef®cient) learning is possible with- can play an important role in the endowment vs. out priori restrictions on the learning space. Sec- learning debate. Linguists' investigations of UG are ond, it is also clear that no matter how much of a rarely developmental, even less so corpus-oriented. head start the child has through UG, language is Developmental psychologists, by contrast, often learned. Phonology, lexicon, and grammar, while stop at identifying components in a cognitive task governed by universal principles and constraints, do [14], without an account of how such components vary from language to language, and they must be work together in an algorithmic manner. On the learned on the basis of linguistic experience. In other hand, if computation is to be of relevance other words±indeed a truism±both endowment and to linguistics, psychology, and cognitive science in learning contribute to language acquisition, the re- general, being merely computational will not suf- sult of which is extremely sophisticated body of ®ce. A model must be psychological plausible, and linguistic knowledge. Consequently, both must be ready to face its implications in the broad empirical taken in account, explicitly, in a theory of language contexts [7]. For example, how does it generalize acquisition [6,7]. to typologically different languages? How does the Controversies arise when it comes to the relative model's behavior compare with that of human lan- contributions by innate knowledge and experience- guage learners and processors? based learning. Some researchers, in particular lin- In this article, we will present a simple compu- guists, approach language acquisition by charac- tational model of word segmentation and some of terizing the scope and limits of innate principles its formal and developmental issues in child lan- of Universal Grammar that govern the world's lan- guage acquisition. Speci®cally we show that SL guage. Others, in particular psychologists, tend to using transitional probabilities cannot reliably seg- emphasize the role of experience and the child's ment words when scaled to a realistic setting (e.g., domain-general learning ability. Such division of child-directed English). To be successful, it must research agenda understandably stems from the di- be constrained by the knowledge of phonological 49 structure. Indeed, the model reveals that SL may CV syllables, many languages, including English, well be an artifact±an impressive one, nonetheless± make use of a far more diverse range of syllabic that plays no role in actual word segmentation in types. And then, syllabi®cation of speech is far human children. from trivial, which (most likely) involve both in- nate knowledge of phonological structures as well 2 Statistics does not Refute UG as discovering language-speci®c instantiations [14]. All these problems have to be solved before SL for It has been suggested [15, 8] that word segmenta- word segmentation can take place. tion from continuous speech may be achieved by using transitional probabilities (TP) between ad- jacent syllables A and B, where , TP(A!B) = 3 The Model P(AB)/P(A), with P(AB) being the frequency of B following A, and P(A) the total frequency of A. To give a precise evaluation of SL in a realis- Word boundaries are postulated at local minima, tic setting, we constructed a series of (embarrass- where the TP is lower than its neighbors. For ex- ingly simple) computational models tested on child- ample, given suf®cient amount of exposure to En- directed English. glish, the learner may establish that, in the four- The learning data consists of a random sam- syllable sequence ªprettybabyº, TP(pre!tty) and TP(ba!by) are both higher than TP(tty!ba): a ple of child-directed English sentences from the CHILDES database [19] The words were then pho- word boundary can be (correctly) postulated. It is remarkable that 8-month-old infants can extract netically transcribed using the Carnegie Mellon Pro- three-syllable words in the continuous speech of an nunciation Dictionary, and were then grouped into syllables. Spaces between words are removed; how- arti®cial language from only two minutes of expo- sure [8]. ever, utterance breaks are available to the modeled learner. Altogether, there are 226,178 words, con- To be effective, a learning algorithm±indeed any sisting of 263,660 syllables. algorithm±must have an appropriate representation of the relevant learning data. We thus need to be Implementing SL-based segmentation is straight- cautious about the interpretation of the success of forward. One ®rst gathers pair-wise TPs from the SL, as the authors themselves note [16]. If any- training data, which are used to identify local min- thing, it seems that the ®ndings strengthen, rather ima and postulate word boundaries in the on-line than weaken, the case for (innate) linguistic knowl- processing of syllable sequences. Scoring is done edge. A classic argument for innateness [4, 5, for each utterance and then averaged. Viewed as an 17] comes from the fact that syntactic operations information retrieval problem, it is customary [20] are de®ned over speci®c types of data structures± to report both precision and recall of the perfor- constituents and phrases±but not over, say, linear mance. strings of words, or numerous other logical possibil- The segmentation results using TP local minima ities. While infants seem to keep track of statistical are remarkably poor, even under the assumption information, any conclusion drawn from such ®nd- that the learner has already syllabi®ed the input per- ings must presuppose children knowing what kind fectly. Precision is 41.6%, and recall is 23.3%; over of statistical information to keep track of. After all, half of the words extracted by the model are not ac- an in®nite range of statistical correlations exists in tual English words, while close to 80% of actual the acoustic input: e.g., What is the probability of a words fail to be extracted. And it is straightfor- syllable rhyming with the next? What is the proba- ward why this is the case. In order for SL to be bility of two adjacent vowels being both nasal? The effective, a TP at an actual word boundary must fact that infants can use SL to segment syllable se- be lower than its neighbors. Obviously, this con- quences at all entails that, at the minimum, they dition cannot be met if the input is a sequence of know the relevant unit of information over which monosyllabic words, for which a space must be pos- correlative statistics is gathered: in this case, it is tulated for every syllable; there are no local min- the syllables, rather than segments, or front vowels. ima to speak of. While the pseudowords in [8] A host of questions then arises. First, How do are uniformly three-syllables long, much of child- they know so? It is quite possible that the primacy directed English consists of sequences of monosyl- of syllables as the basic unit of speech is innately labic words: corpus statistics reveals that on aver- available, as suggested in neonate speech perception age, a monosyllabic word is followed by another studies [18]? Second, where do the syllables come monosyllabic word 85% of time. As long as this from? While the experiments in [8] used uniformly is the case, SL cannot, in principle, work. 50 4 Statistics Needs UG words yields even better performance: a precision of 81.5% and recall of 90.1%. This is not to say that SL cannot be effective for word segmentation. Its application, must be 5 Conclusion constrained±like that of any learning algorithm however powerful±as suggested by formal learning Further work, both experimental and computational, theories [1-3]. The performance improves dramat- will need to address a few pressing questions, in or- ically, in fact, if the learner is equipped with even der to gain a better assessment of the relative contri- a small amount of prior knowledge about phono- bution of SL and UG to language acquisition.
Recommended publications
  • Applications of Universal Grammar (UG) in the ESL/EFL Classroom1
    Applications of Universal Grammar (UG) in the ESL/EFL Classroom1 Lorne O. Kirkwold ABSTRACT The article proposes Stern's (1983) framework for classifying issues related to instruction in order to ascertain the relevance of Universal Grammar (UG) in the ESL/EFL classroom. Discussed in this article, particularly as UG pertains to them, are issues related to: (a) L1transfer; (b) teaching rules and giving error correction versus presenting structures by analogy; and (c) the extent to which the focus should be content rather than language. The article will be of interest especially to teachers and also SLA researchers. The author draws on some examples in English and French, but then presents his conclusion along with further issues he raises based on his recent experience in Japanese universities. INTRODUCTION H. H. Stern will long be remembered as a prominent language educator at the University of Toronto’s prestigious Ontario Institute for Studies in Education for the contribution he made to research in Canada. In his 1983 Fundamental Concepts of Language Teaching, he identified what he perceived as three central issues of language learning: (1) “The L1-L2 connection;” (2) “The explicit - implicit option;” and (3) “The code - communication dilemma” (pp. 400 - 405). These issues as he enumerated them may prove useful as a checklist for classroom teachers for evaluating theoretical framework. In particular, the purpose of this paper is to consider applications of Universal Grammar (UG) which are immediately germane in ESL/EFL instruction. As a definition of UG, Pinker (1995) includes this entry in his glossary: 1 A recent version of this article appears in Studies in Culture, the journal of the Faculty of Humanities, Hokkai- Gakuen University, Sapporo.
    [Show full text]
  • The Growth of Language: Universal Grammar, Experience, and Principles of Computation
    G Model NBR-2698; No. of Pages 18 ARTICLE IN PRESS Neuroscience and Biobehavioral Reviews xxx (2017) xxx–xxx Contents lists available at ScienceDirect Neuroscience and Biobehavioral Reviews journal homepage: www.elsevier.com/locate/neubiorev Review article The growth of language: Universal Grammar, experience, and principles of computation a,∗ b c d e,f Charles Yang , Stephen Crain , Robert C. Berwick , Noam Chomsky , Johan J. Bolhuis a Department of Linguistics and Department of Computer and Information Science, University of Pennsylvania, 619 Williams Hall, Philadelphia, PA 19081, USA b Department of Linguistics, Macquarie University, and ARC Centre of Excellence in Cognition and its Disorders, Sydney, Australia c Department of Electrical Engineering and Computer Science and Department of Brain and Cognitive Sciences, MIT, Cambridge, MA 02139, USA d Department of Linguistics and Philosophy, MIT, Cambridge MA, USA e Cognitive Neurobiology and Helmholtz Institute, Departments of Psychology and Biology, Utrecht University, Utrecht, The Netherlands f Department of Zoology and St. Catharine’s College, University of Cambridge, UK a r t i c l e i n f o a b s t r a c t Article history: Human infants develop language remarkably rapidly and without overt instruction. We argue that the Received 13 September 2016 distinctive ontogenesis of child language arises from the interplay of three factors: domain-specific prin- Received in revised form ciples of language (Universal Grammar), external experience, and properties of non-linguistic domains 10 November 2016 of cognition including general learning mechanisms and principles of efficient computation. We review Accepted 16 December 2016 developmental evidence that children make use of hierarchically composed structures (‘Merge’) from the Available online xxx earliest stages and at all levels of linguistic organization.
    [Show full text]
  • The Problem of Universal Grammar with Multiple Languages: Arabic, English, Russian As Case Study
    (IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 7, No. 4, 2016 The Problem of Universal Grammar with Multiple Languages: Arabic, English, Russian as Case Study Nabeel Imhammed Zanoon Department of Applied Science, Al- Balqa' Applied University (BAU) Aqaba-Jordan Abstract—Every language has its characteristics and rules, to understand and acquire their mother tongue because they though all languages share the same components like words, stipulate that the universal rules are rules for all languages [6]. sentences, subject, verb, object and so on. Nevertheless, Chomsky The language acquisition device is called Universal Grammar suggested the theory of language acquisition in children which provides children with the principles of a universal instinctively through a universal grammar that represents a language and grammatical structures with an instinctive universal grammar for all human languages. Since it has its hypothesis; it suggests that our ability to learn language rules is declaration, this theory has encountered criticism from linguists. already found in the brain. This theory states that language In this paper, the criticism will be presented, and the conclusion ability appears by itself without being taught and that there are that the general rule is not compatible with all human languages characteristics common to all human languages [7]. The will be suggested by studying some human languages as a case problem of the universal grammar with other languages will namely Arabic, English, and Russian. be discussed by studying the syntactic structure in several Keywords—Chomsky; linguistic; Universal Grammar; Arabic; languages and comparing them to each other, and by designing English; Russian a finished cases device that represents the structure and arrangement of words in the languages in order to prove that I.
    [Show full text]
  • Universal Grammar Is Dead 7
    BEHAVIORAL AND BRAIN SCIENCES (2009) 32, 429–492 doi:10.1017/S0140525X0999094X The myth of language universals: Language diversity and its importance for cognitive science Nicholas Evans Department of Linguistics, Research School of Asian and Pacific Studies, Australian National University, ACT 0200, Australia [email protected] http://rspas.anu.edu.au/people/personal/evann_ling.php Stephen C. Levinson Max Planck Institute for Psycholinguistics, Wundtlaan 1, NL-6525 XD Nijmegen, The Netherlands; and Radboud University, Department of Linguistics, Nijmegen, The Netherlands [email protected] http://www.mpi.nl/Members/StephenLevinson Abstract: Talk of linguistic universals has given cognitive scientists the impression that languages are all built to a common pattern. In fact, there are vanishingly few universals of language in the direct sense that all languages exhibit them. Instead, diversity can be found at almost every level of linguistic organization. This fundamentally changes the object of enquiry from a cognitive science perspective. This target article summarizes decades of cross-linguistic work by typologists and descriptive linguists, showing just how few and unprofound the universal characteristics of language are, once we honestly confront the diversity offered to us by the world’s 6,000 to 8,000 languages. After surveying the various uses of “universal,” we illustrate the ways languages vary radically in sound, meaning, and syntactic organization, and then we examine in more detail the core grammatical machinery of recursion, constituency, and grammatical relations. Although there are significant recurrent patterns in organization, these are better explained as stable engineering solutions satisfying multiple design constraints, reflecting both cultural-historical factors and the constraints of human cognition.
    [Show full text]
  • An Overview of Researches on Biolinguistics
    ISSN 1712-8056[Print] Canadian Social Science ISSN 1923-6697[Online] Vol. 10, No. 1, 2014, pp. 171-176 www.cscanada.net DOI:10.3968/j.css.1923669720141001.4219 www.cscanada.org An Overview of Researches on Biolinguistics WU Jieqiong[a],* [a]School of Foreign Languages, Hubei Engineering University, Xiaogan, language to light and promote the further development China. of linguistics. Biolinguistics aims to study languages *Corresponding author. from the perspective of biology and highlights the Received 23 August 2013; accepted 15 January 2014 fact that humanistic linguistics obtains motive force of development from biology (Yuan & Liu, 2008). Abstract Biolinguistics can be interpreted in both narrow and In 1997 in memory of the 40th anniversary of broad senses. Biolinguistics in narrow sense mainly refers transformational-generative grammar, Jenkins wrote to the study on grammar attributes proposed by school the article “Biolinguistics: Structure development and of generative grammar represented by Chomsky, who evolution of language”, which helped produce a large studies languages as a natural object and regards linguistic number of scholarly monographs and papers with respect organic function as human brain’s innate biological organ. to biolinguistics. Simultaneously, a series of relevant In broad sense, the study of biolinguistics means language international academic seminars were successfully held. research from the perspectives of evolutionary biology, This paper, based on the summarization of research neuroscience, genetic science, psychology and even status quo on biolinguistics, looks forward to the future physiological basis for the research of languages (Wu, development of biolinguistics so as to help predict the 2012a). development of biolinguistic researches. Although the discipline of biolinguistics is a new branch of science, it has a long history.
    [Show full text]
  • Another Look at the Universal Grammar Hypothesis: Commentary on Evans 2014 Adele E
    200 LANGUAGE, VOLUME 92, NUMBER 1 (2016) quickly becomes evident that this is not a genuine tour de force, à la Pinker 1994. No, TLM is not the antidote to the long cognitivist nightmare that is The language instinct . In fact, by the end of the book, I fear it not only misses the point, but if adopted to any degree within the field, will also mislead young researchers and students of language into thinking that the debate is between E’s brand of linguistics on the one hand, and a bunch of raving, delusional, irrational bullies on the other. Ultimately, then, TLM will do little to advance the debate on linguistic nativism, although it might very well energize those already in E’s camp. REFERENCES Bickerton, Derek . 1981. The roots of language . Ann Arbor, MI: Karoma. Bickerton, Derek . 1984. The language bioprogram hypothesis. Behavioral and Brain Sciences 7.173–88. DOI: 10.1017/S0140525X00044149 . Crain, Stephen, and Mineharu Nakayama. 1987. Structure dependence in grammar formation. Language 63.522–43. DOI: 10.2307/415004 . Evans, Vyvyan. 2014 . The language myth: Why language is not an instinct . Cambridge: Cambridge Uni - versity Press. Pinker , Steven . 1994. The language instinct . New York: William Morrow & Company. Department of Linguistics University of Hawai‘i at Manoa [[email protected]] Another look at the universal grammar hypothesis: Commentary on Evans 2014 Adele E. Goldberg Princeton University * It is important to recognize that The language myth (TLM ) is not a research monograph, but is instead aimed at a popular audience, and therefore it should be judged in this light.
    [Show full text]
  • The “Grammar” in Universal Grammar: a Biolinguistic Clarification Patrick C
    DOI: 10.1515/qal-2015-0005 ISSN: 2299–8942 The “grammar” in Universal Grammar: A biolinguistic clarification Patrick C. Trettenbrein ABSTRACT This short piece addresses the confusion over terminology that has reigned, and partly still reigns, when it comes to the concept of Universal Grammar (UG). It is argued that whilst there might be changes in terminology and theory, conceptually UG cannot be eliminated. From a biolinguistic perspective, UG is not a hypothesis by any rational epistemological standard, but an axiom. Along these lines, the contemporary evolutionary perspective on the language faculty (FL) is briefly discussed to then argue that UG is necessarily part of FL in both a narrow and broad sense. Ultimately, regardless of terminology, UG is inevitably one of the factors determining the growth of FL. Keywords: Universal Grammar (UG), Faculty of Language (FL), I-language, language universals, minimalism, evolution of language, biolinguistics, cognitive science 1 Introduction Despite having been in active use by (mostly generative) linguists for a long time, the technical term Universal Grammar (UG) has led to ample confusion amongst linguists and non-linguists alike. As a consequence, Boeckx and Benítez Burraco (2014) have supplanted the label UG with talk about the “language-ready brain.” Yet, conceptually, nothing has changed: A language-ready brain is a brain shaped by UG. In turn, UG is defined as genetic endowment with regard to language (Chomsky, 2005) and is one of the factors influencing the initial growth and further development of the human language faculty (FL). This short piece addresses the confusion over the “grammar” in UG because both, terminology and the concept of UG, have been subject to a multitude of debates and misunderstanding.
    [Show full text]
  • How Statistical Learning Can Play Well with Universal Grammar
    Chapter 1 How statistical learning can play well with Universal Grammar Lisa S. Pearl1 1Language Science & Cognitive Sciences, University of California, Irvine, 3151 Social Science Plaza A, Irvine, California, 92697, USA, [email protected] Abstract: One motivation for Universal Grammar (UG) is that it allows children to acquire the linguistic knowledge that they do as quickly as they do from the data that's available to them. A key legacy of Chomsky's work in language acquisition is highlighting the separation between children's hypothesis space of possible representations (typically defined by UG) and how children might navigate that hypothesis space (sometimes defined by UG). While statistical learning is sometimes thought to be at odds with UG, I review how statistical learning can both complement UG and help us refine our ideas about the contents of UG. I first review some cognitively- plausible statistical learning mechanisms that can operate over a predefined hypothesis space, like those UG creates. I then review two sets of examples: (i) those where statistical learning allows more efficient navigation through a UG-defined hypothesis space, and (ii) those where statistical learning can replace prior UG proposals for navigating a hypothesis space, and in turn 1 lead to new ideas about how UG might construct the hypothesis space in the first place. I conclude with a brief discussion of how we might make progress on understanding language acquisition by incorporating statistical learning into our UG-based acquisition theories. Keywords: Universal Grammar, statistical learning, linguistic parameters, parameter setting, linking theories, Subset Principle, syntactic islands 1.1. Introduction A key motivation for Universal Grammar (UG) is developmental: UG can help children acquire the linguistic knowledge that they do as quickly as they do from the data that's available to them [Chomsky, 1981, Jackendoff, 1994, Laurence and Margolis, 2001, Crain and Pietroski, 2002].
    [Show full text]
  • Biolinguistics and the Foundations of a Natural Science of Language*
    teorema Vol. XXXII/2, 2013, pp. 193-204 ISSN: 0210-1602 [BIBLID 0210-1602 (2013) 32:2; pp. 193-204] Biolinguistics and the Foundations of a Natural Science of Language* Cedric Boeckx The Science of Language, Interviews with James McGilvray, by NOAM CHOMSKY, CAMBRIDGE, CAMBRIDGE UNIVERSITY PRESS, 2012, pp. vi+321, £ 15.99. Over the years some of the clearest expositions of Noam Chomsky’s views on language and vision for linguistics have taken the form of inter- views/discussions with experts on language and mind. I have in mind the conversations with Mitsou Ronat that led to Language and Responsibility (1979), or the set of interviews conducted by Henk van Riemsdijk and Riny Huybregts that made The generative enterprise (1982) possible, or the more recent exchange with Adriana Belletti and Luigi Rizzi at the heart of On Na- ture and language (2002). All of these are, in my opinion, core documents to understand Chomsky’s take on language. The science of language, where Noam Chomsky converses with McGill University philosopher of language James McGilvray, is likely to find its niche among the distinguished set of interviews mentioned above, although I should point out right away that I don’t think it reaches the level of its prede- cessors, for reasons I will detail below. The core of the volume is a transcript of “four discussion/interview ses- sions” [p. 3] that took place in 2004, complemented with a follow-up inter- view carried out in January 2009. The interviews are divided into two parts (“The science of language and mind” and “Human nature and its study”), with each one further divided into themes (‘chapters’), although many of the same themes are addressed in multiple chapters.
    [Show full text]
  • Universal Grammar
    Universal Grammar: the history of an idea John Goldsmith March 2, 2010 The term universal grammar first arose (to my knowledge) in 17th century France, in the context of the Port Royal grammarians. But in trying to look back at earlier work, we have to be careful and to bear in mind that until recently, there was no effort made to separate (or even distinguish between) those charac- teristics of language that followed from the nature of thought and/or logic, and those characteristics that do not. A warning about the term deep structure: a term that goes back to Hockett (1958), who wrote of deep grammar. The term was used through Aspects of the Theory of Syntax (Chomsky 1965), and widely misunderstood by people who did not bother to read the book. Deep structure, in Chomsky’s model, has nothing to do with meaning or with what is common across languages. Deep structure corre- sponds to a kind of structurally regular description of a particular language; it is a simplification of surface patterns, and the difference between the observed (surface) patterns and this deep structure are the result of the effect of transfor- mations, which are different in character from the phrase-structure rules which generate deep structures (in this model). phrase-structure rules deep structure of a sentence meaning transformations surface structure of a sentence Katz-Postal-generative semantics hypothesis: deep structures are also meaning representations. Chomsky was unimpressed. Nativism and innatism: in some quar- ters nativism is associated with a sci- 1. UG innatist Hypothesis: Human beings are endowed from birth (if not be- entific reliance on genetics and/or cognitive psychology, while innatism fore!) with some forms of knowledge that did not come from or through goes back to Plato.
    [Show full text]
  • Evans-Levinson BBS 2009.Pdf
    BEHAVIORAL AND BRAIN SCIENCES (2009) 32, 429–492 doi:10.1017/S0140525X0999094X The myth of language universals: Language diversity and its importance for cognitive science Nicholas Evans Department of Linguistics, Research School of Asian and Pacific Studies, Australian National University, ACT 0200, Australia [email protected] http://rspas.anu.edu.au/people/personal/evann_ling.php Stephen C. Levinson Max Planck Institute for Psycholinguistics, Wundtlaan 1, NL-6525 XD Nijmegen, The Netherlands; and Radboud University, Department of Linguistics, Nijmegen, The Netherlands [email protected] http://www.mpi.nl/Members/StephenLevinson Abstract: Talk of linguistic universals has given cognitive scientists the impression that languages are all built to a common pattern. In fact, there are vanishingly few universals of language in the direct sense that all languages exhibit them. Instead, diversity can be found at almost every level of linguistic organization. This fundamentally changes the object of enquiry from a cognitive science perspective. This target article summarizes decades of cross-linguistic work by typologists and descriptive linguists, showing just how few and unprofound the universal characteristics of language are, once we honestly confront the diversity offered to us by the world’s 6,000 to 8,000 languages. After surveying the various uses of “universal,” we illustrate the ways languages vary radically in sound, meaning, and syntactic organization, and then we examine in more detail the core grammatical machinery of recursion, constituency, and grammatical relations. Although there are significant recurrent patterns in organization, these are better explained as stable engineering solutions satisfying multiple design constraints, reflecting both cultural-historical factors and the constraints of human cognition.
    [Show full text]
  • Universal Grammar, Statistics Or Both?
    Opinion TRENDS in Cognitive Sciences Vol.8 No.10 October 2004 Universal Grammar, statistics or both? Charles D. Yang Department of Linguistics and Psychology, Yale University, 370 Temple Street 302, New Haven, CT 06511, USA Recent demonstrations of statistical learning in infants division of labor between endowment and learning: have reinvigorated the innateness versus learning plainly, things that are built in need not be learned, and debate in language acquisition. This article addresses things that can be garnered from experience need not these issues from both computational and developmen- be built in. tal perspectives. First, I argue that statistical learning The influential work of Saffran, Aslin and Newport [8] using transitional probabilities cannot reliably segment on statistical learning (SL) suggests that children are words when scaled to a realistic setting (e.g. child- powerful learners. Very young infants can exploit transi- directed English). To be successful, it must be con- tional probabilities between syllables for the task of word strained by knowledge of phonological structure. Then, segmentation, with only minimal exposure to an artificial turning to the bona fide theory of innateness – the language. Subsequent work has demonstrated SL in other Principles and Parameters framework – I argue that a full domains including artificial grammar, music and vision, explanation of children’s grammar development must as well as SL in other species [9–12]. Therefore, language abandon the domain-specific learning model of trigger- learning is possible as an alternative or addition to the ing, in favor of probabilistic learning mechanisms that innate endowment of linguistic knowledge [13]. might be domain-general but nevertheless operate in This article discusses the endowment versus learning the domain-specific space of syntactic parameters.
    [Show full text]